The midsagittal corpus callosum is topographically organized, that is, with regard to their cortical origin several subtracts can be distinguished within the corpus callosum that belong to specific functional brain networks. Recent diffusion tensor tractography studies have also revealed remarkable interindividual differences in the size and exact localization of these tracts. To examine the functional relevance of interindividual variability in callosal tracts, 17 right-handed male participants underwent structural and diffusion tensor magnetic resonance imaging. Probabilistic tractography was carried out to identify the callosal subregions that interconnect left and right temporal lobe auditory processing areas, and the midsagittal size of this tract was seen as indicator of the (anatomical) strength of this connection. Auditory information transfer was assessed applying an auditory speech perception task with dichotic presentations of consonant−vowel syllables (e.g., /ba-ga/). The frequency of correct left ear reports in this task served as a functional measure of interhemispheric transfer. Statistical analysis showed that a stronger anatomical connection between the superior temporal lobe areas supports a better information transfer. This specific structure−function association in the auditory modality supports the general notion that interindividual differences in callosal topography possess functional relevance.
The corpus callosum is the major fiber tract connecting the 2 cerebral hemispheres, and the axons forming the corpus callosum cross the interhemispheric gap spatially ordered according to their cortical origin (Cipolloni and Pandya 1985; de Lacoste et al. 1985; Schmahmann and Pandya 2006). More specifically, callosal axons from the anterior frontal lobe cross within the most anterior part (genu), whereas pre- and postcentral regions are connected via middle regions (the truncus). Temporal, parietal, and occipital cortices are interconnected via posterior callosal parts (isthmus and splenium). Because the cortical regions are specialized for different functions, this structural organization also establishes a functional topography within the corpus callosum, that is, different callosal subregions are part of different functional networks. This view is supported by clinical studies showing that circumscribed callosal lesions can result in specific functional impairments. For example, interhemispheric transfer of motor and somatosensory information seems to be affected following lesions in middle and posterior truncus (e.g., Risse et al. 1989; Fabri et al. 2001; Caillé et al. 2005), whereas auditory (Sugishita et al. 1995; Pollmann et al. 2002) and visual transfer (e.g., Risse et al. 1989; Corballis et al. 2004) are affected after isthmus and splenium lesions.
The classical approach for the assessment of callosal variability is measuring the midsagittal size and applying a shape-based subdivision procedure to determine callosal subregions (e.g., Witelson 1989; Aboitiz et al. 1992). Using this approach, several studies have successfully demonstrated that interindividual differences in the size of a subregion are associated with various functional indicators of interhemispheric interaction, such as speed of processing (e.g., Schulte et al. 2004; Westerhausen, Kreuder, et al. 2006) and quality of interhemispheric transfer (e.g., Hellige et al. 1998; Westerhausen, Woerner, et al. 2006), interhemispheric interference in a dual-task paradigm (Yazgan et al. 1995), and the synchronization of bilateral movement-related or somatosensory-evoked electroencephalographic activity (Stancak, Hoechstetter, et al. 2002; Stancak, Lucking, and Kristeva-Feige 2002). However, the literature on structure−function associations in callosal subregions is far from consistent, and successful replications are rare. One possible reason for this inconsistency can be found in the huge interindividual variability in the shape of the corpus callosum that hampers the validity of geometrical subdivision procedures (Clarke et al. 1989; Oka et al. 1999). Thus, although gross morphological callosal variability has been shown to be functionally relevant, it can be questioned whether the classical shape-based subdivision methods can capture interindividual differences in specific functional pathways (Denenberg et al. 1991; Hofer et al. 2007). However, with the introduction of the diffusion tensor imaging (DTI) method, it is possible to use noninvasive tractography methods to segment interhemispheric fiber tracts and to map the topography of the corpus callosum (Abe et al. 2004; Huang et al. 2005; Hagmann et al. 2006; Hofer and Frahm 2006; Zarei et al. 2006; Dougherty et al. 2007; Park et al. 2008). DTI tractography relies on the measurement of the diffusion properties of water molecules within the brain parenchyma. As the molecules interact with cellular barriers during the diffusion process, the measured diffusion properties carry information about the microstructural tissue organization (for review, see Le Bihan 2003). Furthermore, myelin sheaths or cell membranes, which are considered the major barriers for water molecules (Beaulieu 2002), hinder the water diffusion stronger perpendicular to than parallel to the longitudinal axis of white matter fiber bundles. As a result, the principle direction of the diffusion process can be seen as an indicator of the main fiber orientation and can thus be used to follow fiber bundles in a 3-dimensional (3D) fashion through the brain (for review see, Nucifora et al. 2007). Thus, the DTI-based tractography offers a unique noninvasive approach to distinguish and study different functional subregions within the corpus callosum.
Applying the DTI tractography to the study of structure and function of the corpus callosum in humans has not only replicated and (mostly) confirmed invasive tracer studies in the macaque brain (Schmahmann and Pandya 2006) but also revealed substantial interindividual differences in the size and location of specific interhemispheric tracts (Abe et al. 2004; Huang et al. 2005; Hagmann et al. 2006; Hofer and Frahm 2006; Zarei et al. 2006; Dougherty et al. 2007; Park et al. 2008). Large interindividual variability is found in particular the posterior corpus callosum, which contains fibers originating from the parietal, temporal, and occipital lobes (cf., Hofer and Frahm 2006; Dougherty et al. 2007). From this observation, it appears reasonable to ask whether such fiber tract variability has functional relevance, for example, whether it is related to efficiency of interhemispheric transfer. However, little is known about the functional relevance of interindividual differences. In a study on reading development in children aged between 7 and 12 years, Dougherty et al. (2007) found a direct link between phonological awareness and the properties of the temporal−callosal pathways that were segmented using DTI tractography. Using interhemispheric inhibition, measured with transcranial magnetic stimulation, Wahl et al. (2007) found a substantial correlation between the microstructural integrity of the motor callosal pathway and inhibition.
The present study was designed to examine the functional relevance of interindividual differences in the architecture of interhemispheric connections in the auditory modality. For this purpose, we used DTI tractography to examine the callosal tracts that interconnect the superior temporal lobe areas (Heschl's gyrus [HG]; posterior superior temporal gyrus [pSTG]) of the left and right hemisphere. We then correlated individual callosal variability with performance on an auditory discrimination task with dichotic presentations of consonant−vowel syllables as a measure of speech perception. This task was chosen because it has been found to be a reliable measure of interhemispheric transfer in the auditory domain (for review see, Tervaniemi and Hugdahl 2003; Bamiou et al. 2007; Westerhausen and Hugdahl 2008). We further predicted that individuals with superior interconnection of the temporal lobe areas would show enhanced reporting of the left ear (LE) stimulus in the dichotic listening task because the LE stimulus has to be transferred across the corpus callosum in order to be processed in the left hemisphere.
Materials and Methods
Seventeen male subjects aged between 19 and 46 years (mean 25.9 ± 6.7 years) participated in the study. All subjects were right handed as assessed by the Edinburgh Handedness Inventory (Oldfield 1971) and had no history of psychiatric or neurological disorders. To ensure normal hearing, all subjects completed audiometric screening for the frequencies 250, 500, 1000, and 2000 Hz (GSI 68; Grason-Stadler Inc., Madison, WI). Only subjects with an auditory threshold below 20 dB and an interaural difference less than 10 dB on any of the tested frequencies were included in the study. All subjects gave their written informed consent prior to participate in the study.
Dichotic Listening Task
The dichotic listening paradigm consisted of pairwise presentations of the 6 constant vowel syllables /ba/, /da/, /ga/, /pa/, /ta/, and /ka/ with 2 different syllables presented on each trial, 1 in the right ear (RE) and 1 in the LE (Hugdahl and Andersson 1986; Hugdahl 2003). All syllables were spoken by an adult Norwegian male voice with constant intensity and intonation. In order to control for systematic effects of syllable voicing (e.g., Berlin et al. 1973; Rimol et al. 2006), only the 12 syllable pairs that result from combining the syllables with equal voicing (e.g., /ba/-/da/, /pa/-/ta/, etc.) were used in the study. The syllables in each pair were temporally aligned to achieve simultaneous onset of the initial consonants, and the mean stimulus duration was approximately 400 ms, with an interstimulus interval of 4000 ms. Each of the 12 syllable pairs was presented 3 times via headphones, resulting in a total of 36 presentations. Stimulus administration and response collection were controlled by the E-Prime software (Psychology Software Tools Inc., Pittsburgh, PA).
The subjects were instructed to report the syllable which they heard best on each trial. The percentage of correctly reported syllables was scored separately for the LE and the RE stimulus, for each individual. A laterality index (LI) was calculated representing the percentage difference between RE and LE scores, from the formula: LI = 100*(RE − LE)/(RE + LE), adapted from Hugdahl and Andersson (1986). Mean correct RE report was 57.4% (±8.3%) and the mean LE report was 37.3% (±6.2%), resulting in a significant RE advantage (t16 = 6.66, P < 0.001) and indicated by a positive LI of 21.0% (±12.8%). Of the 17 subjects, 16 showed a positive LI, whereas one subject had an LI = 0.0%, and none had a negative LI.
Magnetic Resonance Imaging
All magnetic resonance (MR) scans were performed on a 3.0 Tesla GE-Signa system (General Electric Medical Systems, Milwaukee, WI). The scanning protocol consisted of an initial 3D survey scan, followed by a high-resolution anatomical sequence and a DTI sequence. The high-resolution anatomical images were obtained with a T1-weighted pulse sequence (Fast Spoiled Gradient Recall, echo time [TE] = 14 ms, repetition time = 400 ms, inversion time = 500 ms) acquiring 188 consecutive sagittal slices (1 mm thick, scan matrix: 256 × 256) with a field of view of 256 × 256 mm2 and reconstructed to a voxel size of 1 × 1 × 1 mm3. For the DTI measure, a diffusion-weighted single-shot echo planar imagining (EPI) sequence (TE = 89 ms, flip angle = 90°) was used, applying diffusion-sensitizing gradients in 25 directions (weighting factor: b = 1000 s/mm2) and recording 6 reference images (b = 0 s/mm2). In total, 45 axial slices (2.4 mm thickness, no gap, scan matrix: 128 × 128) with a field of view of 220 × 220 mm2, reconstructed to a 128 × 128 matrix, resulting in a voxel resolution of 1.72 × 1.72 × 2.4 mm3. Diffusion tensor modeling and probabilistic tractography were carried out using the FMRIB's Diffusion Toolbox (version 2.0; Behrens et al. 2003, 2007) as implemented in the FSL 4.0 (http://www.fmrib.ox.ac.uk/fsl/). Diffusion weighted images data were first corrected for eddy current distortions. For the evaluation of white matter microstructure, 2 maps of quantitative diffusion parameters were calculated, namely the mean diffusion (MD) and the fractional anisotropy (FA). MD maps were computed as the mean of the 3 eigenvalues in each voxel (with low values indicating low overall diffusion). FA maps were obtained using the equation provided by Pierpaoli and Basser (1996). FA was given preference over other indices of anisotropy because of its lower sensitivity to noise as was demonstrated in several simulation studies (Hasan et al. 2004). The FA values range from 0 for isotropic diffusion to 1 for complete anisotropic diffusion.
Fiber tracts originating from the temporal lobes cross anterior−posterior running association fiber bundles (i.e., inferior and superior occipitofrontal fasciculus) and ventral−dorsal running projection fibers (superior thalamic radiation) via the corpus callosum to the contralateral hemisphere (Wakana et al. 2004). As the directional diffusion information of differently oriented tracts is averaged in voxels located at the fiber crossings, this can in turn result in a deflection of the DTI tractography. Thus, to allow tractography through such regions of fiber crossing, a probabilistic tractography method based on a multifiber model was applied in the present study by utilizing the tractography routines implemented in the FMRIB's Diffusion. This approach estimates in each voxel a probability distribution of each fiber direction and, thus, accounts both for the uncertainty in the data and for the possible existence of multiple fiber directions within each voxel. Based on these local estimates, the global connectivity can than be estimated by following multiple streamline samples through the brain in order to build up an estimate of the distribution of connections from its starting or “seed” voxel. When these streamlines reach a voxel in which more than one direction is probable, they follow the direction that is closest to parallel with the direction at which the streamline arrives (Behrens et al. 2003, 2007).
In the present study, different individualized seed masks were created for each subject, one encompassing HG (as the primary auditory cortex) and one representing the pSTG of each the left and the right hemisphere. The pSTG area was chosen because it has been shown to be involved in acoustic−phonetic and phonological processing (Boatman 2004; Hickok and Poeppel 2007) and has been consistently found to be activated in consonant−vowel dichotic listening paradigms (e.g., Hugdahl et al. 1999; Tervaniemi and Hugdahl 2003; van den Noort et al. 2008). The masks were obtained by separately applying the following steps in each hemisphere. First, based on the digital brain atlas available with the MARINA software package (http://www.bion.de; Walter et al. 2003), a general HG and pSTG was created in standard stereotactic space (Montreal Neurological Institute [MNI]; resolution: 1 × 1 × 1 mm3) in which all voxels anterior to the posterior border of HG were removed (see Fig. 1). In a next step, these masks were individually transformed to the native space of each subject. For this purpose, the individual T1-weighted images were transferred to MNI space and segmented into gray and white matter images using the standard SPM5 algorithms (Wellcome Department of Cognitive Neurology, London, UK).
The normalized HG and pSTG masks were then overlaid on the segmented image, and the gray matter information was used to restrict the masks to include only voxels showing a high gray matter probability. Subsequently, applying a in-house developed algorithm written in MATLAB, the resulting gray matter mask was expanded into white matter tissue by including a neighboring voxel (defined as those 6 voxels sharing a side with a mask voxel) had a white matter probability larger than a predefined threshold (>0.60). This algorithm was applied repeatedly (5 times), resulting in a maximal mask intrusion of 5 mm into the white matter. In a last step, these individualized masks had to be transferred from MNI to native space, as given by the diffusion-weighted images. Due to geometrical distortion in the diffusion-weighted images as opposed to the T1-weighted images, the inverse transformation of the already normalized T1-weighted images is not sufficient enough to reflect the native space of the diffusion-weighted images. Therefore, the back transformation of the masks were achieved by first determining the deformation fields for a nonlinear normalization of b0 reference image of each subject to the standard MNI EPI template. Thereafter, the obtained deformation parameters were inverted using the SPM5 transformation toolbox. These inverted parameters were applied to the HG and pSTG masks for each subject and hemisphere, allowing for an exact inversion of the normalization. In addition to these masks, one mask covering the total corpus callosum on the midsagittal slice (CC mask) was created by manually drawing a region of interest on a b0 image of each subject using the MRIcro software (version 1.40; Rorden and Brett 2000).
For each subject, the tractography algorithm was then applied to calculate 6 different tracts, each connecting a pair of masks: 1) the left and the right HG seed mask, 2) the left and the right pSTG seed mask, 3) the left HG with the CC mask, 4) the right HG with the CC mask, 5) the left pSTG with the CC mask, and 6) the right pSTG with the CC mask. This procedure generates a combined connectivity distribution between the 2 masks and retains only those tracts that pass through both masks. For all calculations, 25 000 streamline samples were drawn, and in order to constrain the estimation, the step length was set to 0.5 mm and a curvature threshold of 0.2 (approximately ±80 degrees) was chosen. The resulting tracts are volumes in which values at each voxel represent the number of samples passing through each voxel and, therefore, the probability of that fibers running through a specific voxel interconnecting voxels within the 2 seed masks. These raw tracts were thresholded (made binary) to remove spurious connections, whereby a probability threshold of 0.001% was applied for the tracts 1 and 2 (i.e., with at least 25 samples passing through a voxel) and 0.005% for the tracts 3−6 (i.e., at least 125 samples passing through a voxel). These relatively low thresholds allow for a high sensitivity to weak paths, what is necessary considering the intrahemispheric pathways that need to be crossed on the way to and from the corpus callosum (Johansen-Berg et al. 2007). Different thresholds were applied in the generation of the 2 first as compared with the tracks 3−6 to compensate for an increased influence of noise and a higher likelihood of deflections due to fiber crossings, which both result from the longer distance between the 2 mask regions (Johansen-Berg et al. 2007).
For further processing, the midsagittal location and size of the resulting tracts were determined by multiplying each of the thresholded tracts (and for each threshold level) with the binary CC mask. For the direct left-to-right HG and pSTG tracts, the resulting midsagittal area was considered to be the crucial corpus callosum region containing the axons interconnecting left and right temporal lobe areas. Applying this “direct tracking approach,” a tract was successfully segmented in only 3 of the 17 subjects (17.6%) for the HG masks. As a consequence, the direct HG tracking approach was removed from further analysis. However, in 11 subjects (64.7%), a direct pSTG connecting tract was found with an average midsagittal tract size of 10.4 voxel (standard deviation [SD]: 14.0 voxel). Figure 2 shows an example for a direct connection for a single subject. Furthermore, in a second approach (combined tracking approach), a combination of the tracts that were calculated connecting the CC mask with the left and right HG or pSTG mask were used for tract segmentation. It should be noted that only interconnections between homotopic regions of both hemispheres were measured. Possible heterotopic connections were not considered because they are known to be weaker and, hence, more difficult to be detected (e.g., Schmahmann and Pandya 2006).
By considering the overlapping area of the 2 tracts as the region of interest, here a substantial tract for the HG was found in 14 subjects (82.4%) and for the pSTG in 15 subjects (88.2%) with a mean midsagittal size of 7.0 voxels (SD: 9.6 voxel) and 24.0 voxel (SD: 18.9 voxel), respectively. Although the tract size significantly diverged between the 2 pSTG tract segmentation approaches (Wilcoxon matched pair test: Z = 3.34, P < 0.001), the results were also highly intercorrelated (Spearman rank r = 0.83; P < 0.001). Taken these observations together, the direct tracking approach can be considered to be a more conservative estimation of tract location. For this reason, both tract size measures were submitted to further analysis.
To assess the microstructural properties of the revealed tracts, the results of the combined tracking approach were also transferred to the FA and MD maps. The median value for both parameters was calculated for those voxels located on the 3 midsagittal slices. The median value was chosen to minimize the effect of outlier voxels (that are most likely due to partial volume effects) on the statistics.
Subdivision of the Corpus Callosum
To compare the tractography-based tract segmentation with traditional approaches for the quantification of callosal variability, we also performed a callosal subdivision based on a standardized geometrical operation. For this purpose, the corpus callosum was first traced on a midsagittal view of the T1-weighted anatomical images as well as on a reference image of the diffusion sequence. This was performed manually for each subject with the software MRIcro. Following a classical approach for morphometrical analysis (see, e.g., Witelson 1989; Aboitiz et al. 1992), the obtained total corpus callosum masks were subdivided into 4 subregions. Referring to a horizontal line connecting the anterior-most and the posterior-most pixels, the genu and the truncus of the corpus callosum were defined as the anterior third and the middle third of the total line length, respectively. The splenium was defined as the posterior fifth and the isthmus as the posterior third minus the posterior fifth. The size of the subregions was automatically extracted. To assess the diffusion properties, the regions of interest were transferred onto the parameter maps and median of MD and FA were calculated for each subregion. To assess reliability, the corpus callosum of each subject was traced a second time by a different rater, yielding intraclass correlations above r = 0.92 for the total corpus callosum area. In light of this high interrater reliability, the mean of the 2 measures was submitted to statistical analysis.
The analyses of the association of corpus callosum measures and dichotic listening performance were grouped into confirmatory and exploratory data analyses, respectively, which were treated differently. The confirmatory part contained correlations relating the dichotic listening performance to the macrostructural corpus callosum measures (i.e., the tract size as determined with the segmentation approaches and the size of the operationally defined subregions). Based on previous findings, it was predicted that a larger callosal area should be related to a superior LE performance and a reduced LI (for review, see Westerhausen and Hugdahl 2008). Because of these a priori hypotheses, one-tailed testing was performed. Because the direct tracking segmentation produced tract estimates that were nonnormally distributed (Shapiro−Wilk test: W = 0.77, P < 0.001), analyses including these measures were performed using Spearman rank correlations (rsp). All other correlations were calculated using Pearson product moment correlations (r). The exploratory part contained the analyses relating the microstructural corpus callosum measures (FA, MD) to the dichotic listening variables as well as analyses relating tracking-based segmentation results to the operationally defined callosal subregions. Here correlations were tested 2 tailed. No adjustment for multiple testing was done in order to retain statistical power (Cohen 1988).
The results of the fiber tracking are shown in Figure 3, revealing the midsagittal location of the different tracts. Relating the tracking results to the defined subregions, the temporal callosal tracts mainly pass the corpus callosum in the splenium region: for the direct tracking approach, 98.3% of the pSTG tract voxels were located in the splenium subregion, whereas it was 94.2% and 92.1% for the for pSTG- and the HG-based tracking, respectively, using the combined tracking approach. All the remaining tract voxels were located in the isthmus, so that the total tract was always located within the posterior third subregion of the corpus callosum. However, for none of the tracking approaches, a significant correlation between tract size and the size of the posterior third region was found (all r2 < 0.10; P > 0.24).
Relating the dichotic listening performance to the results of the tractography-based segmentation, no significant correlation between the midsagittal tract size and LI was found (see Table 1). However, analyzing LE and RE performance separately, significant positive correlations between tract size and the LE report were found for both direct (rsp = 0.57; P < 0.01; r2 = 0.32) and combined (r = 0.54; P = 0.01; r2 = 0.29) tracking approach (see Fig. 4). No such association was found for the RE report. Furthermore, neither HG-based tracking tract measures nor operational defined subregional areas were correlated with the dichotic listening performance (cf., Table 1). Finally, the exploratory analyses did not reveal any significant correlation between microstructural callosal measures and any of the dichotic listening measures.
|Method||Measure||Dichotic listening performance|
|Method||Measure||Dichotic listening performance|
Note: LE/RE correct left and right ear recall, respectively; LI = Laterality index (details see text). (a) Spearman rank correlations.
Statistical probability is indicated as follows: **P < 0.01; *P < 0.05 (all other correlations coefficients r > |0.32| show a trend towards statistical significance; P < 0.10).
The main finding of the present study was that interindividual variability in midsagittal callosal topography was reflected in behavioral outcome. Using a DTI tractography approach to determine the callosal pathways, interconnecting the pSTG of both hemispheres, and relating this to the performance on the dichotic listening paradigm, a significant structure−function association was observed, that is, the LE reports were significantly positively correlated with the midsagittal tract size, whereas the RE reports showed no such association. Moreover, this pattern of association was concordantly found for both the direct and combined tracking approach.
It should be noted that the probabilistic tractography method gives the probability with which a voxel is connected to a defined seed region (Behrens et al. 2003, 2007). As a result, the size of the tract on the midsagittal corpus callosum can be seen as the area that (most likely) contains fibers connecting the speech processing temporal lobe areas. Thus, the positive correlation of tract size and LE recall indicates that stronger connectivity is associated with better recall of the LE stimulus, that is, increased callosal connectivity has functional relevance for reporting of a lateralized speech stimulus. This is in agreement with the structural model of dichotic listening, as originally proposed by Kimura (1967). This model maintains that the RE advantage is caused by the interaction of left hemisphere processing dominance and the anatomy of the auditory projections (see also Sparks and Geschwind 1968; Hugdahl 2003; Westerhausen and Hugdahl 2008). The ascending auditory projections connect each ear to the auditory cortex of both the contra- and the ipsilateral cerebral hemisphere. However, the contralateral projections are more preponderant, resulting in a stronger representation in the hemisphere opposite to the originating ear (cf., Fujiki et al. 2002). The structural model further assumes that the weaker ipsilateral pathways are blocked or inhibited during dichotic stimulation. Because consonant−vowel syllables are processed only in the left hemisphere (cf., Pollmann et al. 2002), the LE stimulus—initially transmitted to the right hemisphere—needs to be transferred via the corpus callosum to the left hemisphere in order to get processed. The RE stimulus, on the other hand, is transmitted directly to the left hemisphere and needs not to be transferred across the corpus callosum. The inferior LE report in the dichotic listening task is therefore explained by a delay and/or attenuation of information related to the additional callosal signal transfer. A callosal relay hypothesis is also supported by clinical studies showing that an acquired (permanent or transient) loss or degradation of callosal connections goes along with reduced LE report in dichotic listening (e.g., Reinvang et al. 1994; Benavidez et al. 1999; Gadea et al. 2002; Pollmann et al. 2002; Mataro et al. 2006). Following a complete callosal section, dichotic listening performance is characterized by significantly reduced reports for the LE stimulus, which are usually not above chance level (e.g., Milner et al. 1968; Sparks and Geschwind 1968; Springer and Gazzaniga 1975; Clarke et al. 1993). The present results extend the clinical findings by demonstrating that not only the complete destruction of the corpus callosum but also interindividual variability in temporal−callosal pathways is predictive of performance on the dichotic listening task.
In the present study, probabilistic DTI tractography was used to assess the interindividual variability in the callosal pathways originating from the superior temporal lobe areas (HG, pSTG) and, thus, in an anatomical and functional specific callosal tract. In showing the functional relevance of such specific callosal pathways, the present study extends previous findings, which have not distinguished between the different pathways within the corpus callosum (e.g., Hines et al. 1992; Clarke et al. 1993; Yazgan et al. 1995; Gootjes et al. 2006; Westerhausen, Woerner, et al. 2006). The present findings also supplement previous results that have examined other functional pathways (Wahl et al. 2007) or using divergent functional approaches (Dougherty et al. 2007). Furthermore, a tractography-based segmentation of a specific tract can be expected to be a more sensitive method to assess functionally relevant interindividual differences (e.g., Dougherty et al. 2007; Westerhausen and Hugdahl 2008). Accordingly, although finding significant structure−function correlations for the tractography-derived callosal measures, no such association was found relating conventional cross-sectional area measures of callosal size to dichotic listening performance in the present study. Moreover, tractography-derived track sizes were not significantly correlated to the midsagittal area measures of the posterior callosal regions, indicating low predictive value of gross anatomical measures for differences in (functionally) specific callosal pathways (at least for the examined temporal−callosal auditory pathways). The present study supports the notion that DTI tractography−based tract segmentation might serve as a reliable method to assess functionally relevant interindividual callosal tract variability in future studies.
The present study also reveals a better localization of the auditory pathways within the corpus callosum. As can be seen in Figure 3, the tracts connecting the HG regions in the right and left hemisphere were found to be located more rostrally within the posterior callosal third than the pSTG connecting tracts. This seems to fit well with the general topographical pattern that callosal terminals originating from more rostrally situated temporal cortical areas also cross the corpus callosum more rostrally (Cipolloni and Pandya 1985). Earlier observations in the rhesus monkey, with autoradiographic techniques also indicate that the superior temporal lobe regions are interconnected through callosal areas located at the intersection of the isthmus and splenium regions (Cipolloni and Pandya 1985; Schmahmann and Pandya 2006). However, a more caudal tract location was first indicated by clinical studies of circumscribed callosal lesions, pointing at a critical role of the splenium (rather than the isthmus) for the transfer of auditory information (Sugishita et al. 1995; Pollmann et al. 2002). Pollmann et al. (2002) showed that a lesion in the posterior 20% of the corpus callosum resulted in suppression of reporting of the LE stimulus in both a dichotic monitoring (target detection) and a standard (free report) dichotic listening task, whereas a lesion in more rostral regions did not cause a noticeable deficit in LE stimulus report. This observation has more recently been supported by DTI tractography studies that have found that the temporal−callosal pathways are located predominately in the splenium region (e.g., Hagmann et al. 2006; Hofer and Frahm 2006; Zarei et al. 2006; Park et al. 2008). The results of the present study further support the importance of the splenium region. Despite interindividual variability in size and exact location, the tracts connecting the pSTG mainly runs through the splenium (defined as the posterior fifth) of the corpus callosum (see Fig. 3), what is supported by the fact that on average 92.1−98.3% (depending on the tract segmentation procedure) of the tract voxels were found to be located in the splenium. However, extending beyond previous findings, the present study also revealed a significant association between the segmented tract and indices of interhemispheric interaction (i.e., dichotic listening) that supports the functional importance of this region.
In summary, the present study supports the notion that DTI tractography can be used to map different pathways on the midsagittal corpus callosum and study their functional relevance. By focusing on the auditory domain, it was shown that interindividual differences in the strength of the temporal−callosal connections are associated with differences in the interhemispheric interaction in a dichotic listening paradigm. However, the cytoarchitectonical basis and the exact mechanisms underlying this interaction need to be clarified in future studies. Histological studies have shown that the corpus callosum is composed of axons that largely differ in myelin thickness and diameter (LaMantia and Rakic 1990; Aboitiz et al. 1992). Thus, the variability in interhemispheric connectivity might reflect differences not only in the number of axonal projections but also in the myelination or size of the axons. The role of the callosal axons might be found in the synchronization or desynchronization of neuronal activity in both hemispheres (Aboitiz et al. 2003; Brancucci et al. 2008) what can be seen as a key mechanism underlying the dynamical coordination and integration of (interhemispherically) distributed brain processes (e.g., Varela et al. 2001).
Faculty of Psychology, University of Bergen, Norway (to K.H.).
The authors like to thank Roger Barndon and Helene Hjelmervik for their valuable support in the data collection. Conflict of Interest: None declared.